Design and Development of a Personal Protective Equipment (PPE) Detection Application for Construction Workers Using The YOLO Algorithm
Abstrak
Occupational Health and Safety (OHS) is an essential aspect of the construction sector. However, low compliance among workers in wearing Personal Protective Equipment (PPE) remains a major cause of workplace accidents. This study aims to design a real-time PPE detection system using the You Only Look Once (YOLO) algorithm. A dataset of 1,200 images with 14,819 bounding boxes was collected and trained using YOLOv8, achieving 95.4% precision, 94.9% recall, and 95.9% mAP@50. The system was implemented as a web-based application using Flask, supporting both image uploads and RTSP CCTV streaming. A filtering feature was included to help supervisors focus on specific violations such as “no helmet” or “no vest.” Testing demonstrated near real-time performance of 16–18 FPS on 720p video input. The system effectively improves OHS monitoring and can be further enhanced through model optimization and automated alert features.